Meituan-Robotics-0: A Vision-Language-Action Foundation Model for Desktop Manipulation
Meituan-Robotics-0: A Vision-Language-Action Foundation Model for Desktop Manipulation
This paper presents Meituan-Robotics-0 (MR0), a vision-language-action foundation model designed for stationary-base desktop manipulation.
The paper introduces Meituan-Robotics-0 (MR0), a vision-language-action foundation model for stationary-base desktop manipulation, trained on 147 million samples and 15,060 hours of trajectories across 12 embodiments, achieving 93.6% success on RoboTwin 2.0 and fifth place in the RoboDojo VLA track.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
vision-language samples
147,000,000
View original evidence
arXiv:2611.08577v1 Announce Type: new Abstract: We present Meituan-Robotics-0 (MR0), a vision-language-action foundation model and embodied-agent system for stationary-base desktop manipulation. MR0 combines 147 million vision-language samples with 15,060 hours of curated robot trajectories spanning 12 embodiments. A three-stage training recipe aligns multimodal representations with robot behavior through discrete ac
Open source S2robot trajectories hours
15,060 hours
View original evidence
arXiv:2611.08577v1 Announce Type: new Abstract: We present Meituan-Robotics-0 (MR0), a vision-language-action foundation model and embodied-agent system for stationary-base desktop manipulation. MR0 combines 147 million vision-language samples with 15,060 hours of curated robot trajectories spanning 12 embodiments. A three-stage training recipe aligns multimodal representations with robot behavior through discrete ac
Open source S2progress score Ethernet insertion
0.9
View original evidence
MR0 achieves higher average success rate and progress score than $\pi_{0.5}$ across three real-robot tasks. On Ethernet insertion, CFG-DAgger trained with Rollback data reaches 70% success and a progress score of 0.90. Qualitative agent deployments demonstrate cross-task event retrieval and recovery from temporarily unsatisfied task conditions.
Open source S5
Dataset details reported by the source
- Modalities
- vision-language-action Open source S2
- Embodiments
- 12 embodiments Open source S2
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
What remains unknown
Dataset licensing and commercial-use rights are not established here.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
arXiv:2611.08577v1 Announce Type: new Abstract: We present Meituan-Robotics-0 (MR0), a vision-language-action foundation model and embodied-agent system for stationary-base desktop manipulation. MR0 combines 147 million vision-language samples with 15,060 hours of curated robot trajectories spanning 12 embodiments. A three-stage training recipe aligns multimodal representations with robot behavior through discrete ac
Open source S2
ous embodiments, while asynchronous action-chunk merging maintains smooth execution under variable inference latency. An event-driven agent maintains persistent task events and reassesses plans when progress or feasibility changes, enabling cross-task retrieval and conditional task resumption. MR0 achieves 93.6% success on RoboTwin 2.0 and ranks fifth in the RoboDojo VLA track. Under matched HG-DAgger post-training,
Open source S4
MR0 achieves higher average success rate and progress score than $\pi_{0.5}$ across three real-robot tasks. On Ethernet insertion, CFG-DAgger trained with Rollback data reaches 70% success and a progress score of 0.90. Qualitative agent deployments demonstrate cross-task event retrieval and recovery from temporarily unsatisfied task conditions.
Open source S5
Implications for data suppliers
RoboSignal interpretation and collection questions, not statements of buyer demand.
- Confirm the required data type and collection setting with the buyer; this source does not establish a complete collection specification.
- Compare the reported units and scope before using these quantities in a budget. Recording hours, sensor-hours and trajectories are different measures.
- Check the exact dataset version, license, modalities and task coverage against the intended use.
- Validate demand and acceptance criteria with a buyer before scaling. Publication, popularity and a research result do not establish a purchase commitment.
Source：arXiv Robotics — research abstracts · arxiv.org